The use of proper child restraint systems (CRS) is mandatory for children travelling in cars in most countries of the world. The analysis of the quantity of restrained children shows that more than 90% of the children in Germany are restrained. Looking at the quality of the protection, a large discrepancy between restrained and well protected children can be seen. Two out of three children in Germany are not properly restrained. In addition, considerable difference exists with respect to the technical performance of CRS. For that reason investigations and optimisations on two different topics are necessary: The technical improvement of CRS and the ease of use of CRS. Consideration of the knowledge gained by the comparison of different CRS in crash tests would lead to some improvements of the CRS. But improvement of child safety is not only a technical issue. People should use CRS in the correct way. Misuse and incorrect handling could lead to less safety than correct usage of a poor CRS. For that reason new technical issues are necessary to improve the child safety AND the ease of use. Only the combination of both parts can significantly increase child safety. For the assessment of the safety level of common CRS, frontal and lateral sled tests simulating different severity levels were conducted comparing pairs of CRS which were felt to be good and CRS which were felt to be poor. The safety of some CRS is currently at a high level. All well known products were not damaged in the performed tests. The performance of non-branded CRS was mostly worse than that of the well known products. Although the branded child restraint systems already show a high safety level it is still possible to further improve their technical performance as demonstrated with a baby shell and a harness type CRS.
Since a number of human models have been developed it appears sensible to use these models also in the accident analysis. Especially the understanding of injury mechanisms and probably even injury risk curves can be significantly improved when interesting accidents are reconstructed using human body models. However, an important limitation for utilising human models for accident reconstruction is the effort needed to develop detailed FE models of the accident partners or to prepare the human model reconstruction by running physical accident reconstructions. The proposed approach for using human models for accident reconstruction is to use simplified and parametric car models. These models can be adapted to the crash opponents in a fast and cost effective way. Although, accuracy is less compared to detailed FE models, the relevant change in velocity can be simulated well, indicating that the computation of a detailed crash pulse is not needed. Two frontal impact test accidents that were reconstructed experimentally and using the parametric car models are indicating sufficient correlation of the adapted parametric car models with the full scale crash reconstructions. However, further developments of the parametric models to be capable for the use in lateral impacts and rear impacts are needed. For the PC Crash simulation runs the output sampling rate is too large to allow sufficient analysis. In addition the performance appears to be too general.